phi3-offline-dpo-lora-noise-0.0-5e-7-42

This model is a fine-tuned version of microsoft/Phi-3-mini-4k-instruct on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.6914
• Rewards/chosen: -0.0098
• Rewards/rejected: -0.0147
• Rewards/accuracies: 0.5913
• Rewards/margins: 0.0049
• Logps/rejected: -385.2070
• Logps/chosen: -409.5497
• Logits/rejected: 12.4993
• Logits/chosen: 14.2906

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-07
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• total_eval_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.6933	0.1778	100	0.6932	0.0002	-0.0009	0.5278	0.0011	-383.8265	-408.5499	12.5336	14.3223
0.6942	0.3556	200	0.6928	-0.0035	-0.0064	0.6111	0.0029	-384.3740	-408.9154	12.5221	14.3092
0.6933	0.5333	300	0.6918	-0.0073	-0.0121	0.5794	0.0048	-384.9481	-409.3039	12.5135	14.3020
0.6928	0.7111	400	0.6914	-0.0104	-0.0142	0.5714	0.0038	-385.1545	-409.6051	12.4934	14.2847
0.6935	0.8889	500	0.6912	-0.0101	-0.0159	0.5913	0.0058	-385.3226	-409.5751	12.4995	14.2891

Framework versions

• PEFT 0.7.1
• Transformers 4.42.3
• Pytorch 2.3.0+cu121
• Datasets 2.14.6
• Tokenizers 0.19.1